﻿ 基于多目标遗传算法的舰载液冷板散热优化设计
 舰船科学技术  2018, Vol. 40 Issue (6): 134-138 PDF

Optimization design of liquid marine cooling plate based on multiple objective genetic algorithm
YANG Ping, CHEN Zheng-jiang, HUANG Wei
The 30 Research Institute of CETC, Chengdu 610041, China
Abstract: Aiming at the problem of marine cooling plate heat exchange and pressure loss, Using the variable cross-section channel design to improve the Reynolds coefficient, in order to achieve the goal of efficient heat exchanger . building functional relationships of liquid cooling plate by Using BP network, In order to improve the computational efficiency. With the highest temperature of key device and cooling plate circulation pressure as the design object, and the geometry size of the fin section in channel and the length of variable cross-section as design variables, No-dominated Sorting Genetic Algorithm with punishment operator is used to obtain Pareto front solution of the cooling system. According to the actual use environment, a set of solutions is selected for engineering verification,The using of this optimization method has an important guiding significance for design of cooling plate.
Key words: liquid cooling plate     heat exchange     BP network     NSGA-2 algorithm     pareto front
0 引　言

1 液冷板模型与仿真计算 1.1 液冷模型

 图 1 冷板流道外形图 Fig. 1 Appearance of cold plate flow-path
1.2 仿真计算

 $\varPhi = {q_m}.{c_p}.({t_2} - {t_1})\text{。}$

 $\Delta P = f\frac{{L{V^2}}}{{de^2g}}{10^{ - 4}}\text{，}$

 $V = \frac{Q}{{A\rho }}\text{，}$
 $f = \frac{{0.316}}{{{{Re}^{0.25}}}}\text{。}$

 图 2 射频阵列温度与冷板流场压力图 Fig. 2 Temperature of Radio frequency array and stress of cold plate flow-path
2 基于BP网络的映射函数

 图 3 神经网络拓扑图 Fig. 3 Neural network topology

 ${net}_{{i}}^{{p}} = \sum\limits_{j = 1}^n {{w_{ij}}x_j^p - {\theta _i}}\text{，}$
 $O_i^p = g(net_{{i}}^{{p}})\text{。}$

 $\begin{gathered} \Delta {w_{ki}} = - \eta \frac{{\partial {J_p}}}{{\partial {w_{ki}}}} = - \eta \frac{{\partial {J_p}}}{{\partial net_k^p}}.\frac{\partial }{{\partial {w_{ki}}}}\left( {\sum\limits_{i = 1}^q {{w_{ki}}o_i^p - {\theta _k}} } \right) \text{，} \hfill \\ \Delta {w_{ki}} = - \eta \frac{{\partial {J_p}}}{{\partial net_k^p}}.o_i^p\text{。} \hfill \\ \end{gathered}$

 $\mathop n\nolimits_l = \sqrt {n + m} + a{\text{。}}$

 图 4 翘片设计变量图 Fig. 4 Design variable of heat sink

 图 5 BP网络训练收敛曲线 Fig. 5 Convergence curve of BP network
3 优化计算 3.1 设计变量与优化目标

 $T = {f_t}(a,b,c) < 70\text{，}$
 $P = \left| {{f_p}(a,b,c) - 1800} \right| < 225\text{。}$

3.2 算法实施

1）引入精英概念，避免了父代优秀个体流失；

2）使用快速无支配排序法，使种群中的每个个体对应其无支配程度的序号，初步标识了个体的优劣；

3）在同级种群中进行拥挤度计算，保证种群进化的同时兼顾了多样性。

 图 6 算法流程图 Fig. 6 Algorithm flowchart

 $P{[i]_{\rm distance}} = \sum\limits_{k = 1}^r {(P[i + 1]} {f_k} - P[i - 1]{f_k}){\text{，}}$

3.3 计算结果与分析

 图 7 Pareto前沿分布图 Fig. 7 Pareto front

 图 8 翘片宽度与功放最高温度关系图 Fig. 8 Relationship between width of heat fin and max temperature of amplifier

 图 9 翘片高度与冷板压差关系图 Fig. 9 Relationship between height of heat fin and pressure of cold plate
4 结　语

 [1] 吕永超, 杨双根. 电子设备热分析热设计及热测试技术综述及最新进展[J]. 电子机械工程, 2007, 23(1): 5–10. http://www.cqvip.com/QK/93346X/200701/24003520.html [2] 雷俊禧, 朱冬生, 王长宏, 等. 电子芯片液体冷却技术研究进展[J]. 科学技术与工程, 2008, 8(15): 4258–4263, 4269. [3] 周海峰, 邱颖霞, 鞠金山. 电子设备液冷技术研究进展[J]. 电子机械工程, 2016, 32(4): 7–10,15 [4] 薛梅. 车载雷达户外液冷机箱设计与研究[C]//中国电子学会电子机械工程分会, 中国电子学会电子机械工程分会2009年机械电子学学术会议论文集. 2009: 5. [5] 解金华、邹吾松, 等. 某机载S形深孔液冷板优化设计[J]. 电子机械工程, 2014, 30(4): 1–4. http://www.cqvip.com/QK/93346X/201404/662165277.html [6] 尹本浩、蒋威威, 等. 液冷电子设备的冷板流阻匹配研究[J]. 电子机械工程, 2013, 29(2): 1–4. http://www.cqvip.com/QK/93346X/201302/45697465.html [7] 李松, 王皓, 等. 微通道高热流换热分析研究[J]. 机械科学与技术. 2012, 30(3): 384–387. http://cdmd.cnki.com.cn/Article/CDMD-10358-1011125077.htm [8] 田立新, 文尚胜, 等. 大功率LED液冷热沉结构与换热效果研究[J]. 光学学报. 2015, 35(3): 1–8. http://www.opticsjournal.net/abstract.htm?id=OJ150204000033qWtZ3y [9] 余建祖, 高红霞, 等. 电子设备热设计及分析技术[M]. 北京: 北京航空航天大学出版社, 2008. [10] RAJEEV S, KRISNAMOORTHY CS. Genetic algorithm-based methodology for design optimization of reinforced concrete frames. Computer-aided civil and infrasetrucer engineering, 1999(20). 208–212. [11] 闵惜琳, 刘国华. 人工神经网络结合遗传算法在建模和优化中的应用[J]. 计算机应用研究, 2002(1): 79–80. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=jsjyyyj200201030 [12] DEB K, AARAWAL S, PRATAP A, et al. A fast elitist non-dominated sorting genetic algorithm for multi-objective optimization: NSGA-Ⅱ[C]//Proc of the Parallel Problem Solving form Nature Ⅵ Con. Paria, 2000. 849–858.